Monitoring Effluent Water Quality in Wastewater Treatment Process Based on Multi-model Online Soft Sensing Method

For the biochemical reaction stage of the wastewater treatment process, which involves a complex microbial reaction process, there are many parameters that are difficult to monitor online. Therefore, using a multi-model online soft measurement method is a key but challenging problem. In this paper, advanced machine learning technology is applied, and a multi-model online soft sensing method is adopted to address the difficulty of online monitoring of effluent quality parameters in the sewage treatment process. Specifically, the method uses an online subtraction clustering algorithm to divide real-time working condition data samples, and according to the distribution of real-time working condition data samples in the space, the fuzzy strategy is adopted to assign the real-time working condition data of the corresponding subspace to different submodels for learning. Finally, the final soft measurement results are obtained by dynamically integrating the outputs of each submodel. The actual operation data of a sewage treatment plant were used to detect the ammonia nitrogen in the effluent water of the sewage treatment process. Experimental results show that the proposed method is able to self-organize multi-model soft sensor model driven by real-time condition data, and the multi-model soft sensor model constructed by this method has improved the modeling accuracy, modeling speed, and the ability to track real-time conditions.